Cointegration-Based Pair Trading on Brazil’s B3 Market
Summary
This project tests a market-neutral pairs strategy on Brazilian equities, grouping stocks by sector and screening pairs with the Johansen cointegration test. It keeps pairs with a consistently signed spread and a half-life no longer than 60 days. Entry and exit signals use one-standard-deviation Bollinger bands around the spread, with the band length set to the pair’s estimated half-life. An Augmented Dickey-Fuller test is rerun periodically to check whether mean reversion persists; its frequency and lookback are selected by grid search.
The study uses daily closing prices across two historical periods, with the final year held out for testing and the top training performers carried forward. It reports results at several leverage levels and also examines a recent large-cap subset to address short availability. The authors flag survivorship bias in the earlier sample and note that the results omit transaction fees and slippage. Daily closes, limited short liquidity, and pair selection based on historical relationships constrain how directly the reported performance translates to live trading.
Key ideas
- Sector grouping narrows the universe of candidate Brazilian equity pairs.
- The strategy screens for cointegration, stable spread sign, and a bounded mean-reversion half-life.
- Spread deviations trigger positions, while a return to the moving average closes them.
- Periodic ADF checks are tuned to monitor whether the spread remains mean reverting.
- Reported tests omit fees and slippage, and the earlier sample has survivorship bias.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.